Evaluating the Role of Noise Mitigation in QAOA: A Study of Zero-Noise Extrapolation and CVaR
Adriano Lusso, M. Vénere, Victor Onofre, Alberto Maldonado-Romo, J. A. Montañez-Barrera, Marco Domenico Santambrogio · 2025
In recent years, the Quantum Approximate Optimization Algorithm (QAOA) has attracted attention due to its potential to find good-quality solutions to combinatorial optimization problems. In QAOA, there is a set of parameters that are usually tuned using classical optimization methods. These methods search for parameters that identify good solutions to the given problems. Yet, the effects of noise on current Noisy Intermediate-Scale Quantum (NISQ) devices proved to have a destructive impact on the search of such parameters, reducing the quality of the exploration and producing inefficient solutions. In this work, we explore the Zero-Noise Extrapolation (ZNE) technique to mitigate the effects of noise in QAOA, the modification of the objective function using the Conditional Value at Risk (CVaR), and present simulations using both methods on different MaxCut instances. We found that ZNE, in combination with CVar, slightly restores the noisy energy landscape of QAOA, improving the optimal energy found. However, this is not the case when measuring high-quality solutions. While CVaR slightly improves solution quality, ZNE interferes with these results. Finally, we provide an open-source integration of ZNE and CVar, to support researchers in assessing the performance of their QAOA workflows in noisy environments, towards effective applications of QAOA.